Key Statistical Concepts in Clinical Trials for Pharma.pdf

SpringerBriefs in PharmaceuticalScience&Drug DevelopmentFor further volumes:http:/ TurnerKey Statistical Concepts inClinical Trials for Pharma123J.Rick Turner,Ph.D.Cardiac Safety Services,QuintilesEmperor Boulevard 4820Durham,NC 27703USAe-mail:ISSN 1864-8118e-ISSN 1864-8126ISBN 978-1-4614-1661-6e-ISBN 978-1-4614-1662-3DOI 10.1007/978-1-4614-1662-3Springer New York Dordrecht Heidelberg LondonLibrary of Congress Control Number:2011938148?The Author(s)2012All rights reserved.This work may not be translated or copied in whole or in part without the writtenpermission of the publisher(Springer Science+Business Media,LLC,233 Spring Street,New York,NY10013,USA),except for brief excerpts in connection with reviews or scholarly analysis.Use inconnection with any form of information storage and retrieval,electronic adaptation,computersoftware,or by similar or dissimilar methodology now known or hereafter developed is forbidden.The use in this publication of trade names,trademarks,service marks,and similar terms,even if they arenot identified as such,is not to be taken as an expression of opinion as to whether or not they are subjectto proprietary rights.Printed on acid-free paperSpringer is part of Springer Science+Business Media()PrefaceThe ultimate purpose of the results from a clinical trial is not to tell us preciselywhat happened in that trial,but,in combination with results from other trials in thedrugs clinical development program,to gain insight into likely drug responses inpatients who would be prescribed the drug should it be approved for marketing.The discipline of Statistics enables us to do this.This book discusses key statistical concepts that facilitate the analysis of datacollected from a group of individuals participating in a biopharmaceutical clinicaltrial,the estimation of their clinical significance in the general population ofindividuals likely to be prescribed the drug if approved,and the related decisionmaking that occurs at both the public health level(by regulatory agencies whendeciding whether or not to approve a new drug for marketing)and the individualpatient level(by physicians and their patients when deciding whether or not thepatient should be prescribed a drug that is on the market).These key concepts include drug safety and efficacy,clinical significance,statistical significance,and benefit-risk estimation.All of these facilitate decisionmaking during drug development,and also during pharmacotherapy once the drugis marketed.vContents1Setting the Scene.11.1Introduction.11.2The Discipline of Statistics.21.3Generalizing Information Gained from a Clinical Trial.31.4Blood Pressure.41.4.1The Physiology and Measurement of Blood Pressure.51.4.2A Cautionary Tale About Blood PressureMeasurement.61.5Statistical Concepts and Nomenclature.71.5.1Drug Safety.71.5.2Drug Efficacy.81.5.3Statistical Significance and Clinical Significance.8References.102Analyzing Safety Data.112.1Introduction.112.2Providing Safety Data to Prescribing Physicians and Patients.122.3The Clinical Study Report.122.4Subject Demographics and Accountability.132.5Safety Parameters Measured.142.6Adverse Events.142.7From Descriptive Statistics to Inferential Statistics.152.8Prospective Exclusion of Unacceptable Risk.152.8.1Assessment of Unacceptable Cardiac Risk.162.8.2Assessment of Unacceptable Cardiovascular Risk.172.8.3A More Realistic Scenario and itsUnintended Consequences.20References.21vii3Assessing Efficacy Data.233.1Introduction.233.2Probability.243.3Statistically Significant Efficacy and Hypothesis Testing.253.3.1A Straightforward Example.263.3.2A Quick Detour:Degrees of Freedom.273.3.3Returning to Our Example.283.3.4A Second Method of Analysis:ANOVA.293.4Clinically Significant Efficacy and Confidence Intervals.303.5Emphasizing an Earlier Point.31References.324Confidence Intervals:Additional Commentary.334.1Introduction.334.2The Logic of Confidence Intervals.334.3Differing Confidence Intervals Differ in Width.344.4Employment of Confidence Intervals in Both Safetyand Efficacy Analyses.354.4.1Equidistant and Non-Equidistant Confidence Intervals.354.5The Limit of Primary Interest.364.6Relationship Between Confidence Intervalsand Probability Levels.374.6.1Cases in Which the Null Value is Zero.384.6.2Cases in Which the Null Value is Unity.395Meta-Methodology.415.1Introduction.415.2Unbridled Bravado is Inappropriate When Presenting Results.425.3Fundamentals of Meta-Methodology.435.4Data Analysis:Fixed-Effect and Random-Effect Models.455.5Evaluating Heterogeneity.465.6Evaluating Robustness.475.7Hypothesis Generation and Hypothesis Testing.485.8Results and Conclusions.49References.496BenefitRisk Estimation.516.1Introduction.516.2Drug Safety.526.3Decision Making.536.4The Subjective Nature of Many Decisions.536.5Determining and Enforcing Thresholdsof Regulatory Concern.55viiiContents6.6Decision Analysis and Decision Making.566.7Qualitative and Quantitative Considerations.57References.57About the Author.59Index.61ContentsixChapter 1Setting the SceneAbstract The ultimate purpose of the results from a clinical trial is not to tell usprecisely what happened in that trial,but,in combination with results from othertrials in the drugs clinical development program,to gain insight into likely drugresponses in patients who would be prescribed the drug should it subsequently beapproved for marketing.The discipline of Statistics,which can be meaningfullyregarded as a way of conducting business during new drug development,providesthe structural architecture within which all stakeholders operate,and enables theprovision of compelling evidence of safety and efficacy by using procedures thateveryone has agreed to honor.Not everyone needs to be an expert in Statistics,buteveryone engaged in drug development can benefit considerably from a solidconceptual understanding of drug safety and efficacy,clinical significance,sta-tistical significance,and benefit-risk assessments.Keywords Statistical concepts?Clinical trials?Drug safety?Drug efficacy?Statistical significance?Clinical significance1.1 IntroductionThe discipline of Statistics is a central component of drug development.It is truethat not everyone needs to be an expert in Statistics,but the discipline is of suchfundamental importance in the decision-making processes that occur within drugdevelopment that everyone engaged in these endeavors can benefit considerablyfrom a solid conceptual understanding of drug safety and efficacy,clinical sig-nificance,statistical significance,and benefit-risk assessments.Once you have agood understanding of these statistical concepts you will realize that everyoneinvolved in clinical trials needs to play their role in the collection of optimumJ.R.Turner,Key Statistical Concepts in Clinical Trials for Pharma,SpringerBriefs in Pharmaceutical Science&Drug Development,DOI:10.1007/978-1-4614-1662-3_1,?The Author(s)20121quality data in each and every trial.Optimum quality data lead to optimum qualityinformation and optimum quality decisions throughout the drug developmentprocess.The goal of this book is to guide you to this realization while presentingthe absolute minimum amount of mathematical formulas and calculations.Statistics can be meaningfully regarded as a way of doing business.Calcula-tions certainly need to be performed during the statistical analysis of clinical trialdata,but computers are supremely placed to do that part of the overall process:The contribution of professional statisticians is much richer than simply numbercrunching.Statistics provides the cornerstone of designing clinical trials capable ofgenerating optimum quality data whose analysis forms the rational basis ofdecisions.It provides the structural architecture within which all stakeholdersoperate,and enables the provision of compelling evidence by using proceduresthat everyone has agreed to honor.Ethical conduct is absolutely critical at all stages of clinical research.It isappropriate to remind ourselves frequently that our work has a real impact onpatients lives.New drug development is a complicated and difficult undertaking,but one that makes an enormous difference to the health of people worldwide.It isa noble pursuit,and a privilege to be involved in this work.However,with thisprivilege comes the responsibility to conduct our tasks with constant due diligenceand to the highest ethical standards.Derenzo and Moss 1 captured the impor-tance of ethical considerations in all aspects of clinical studies in the followingquote:Each study component has an ethical aspect.The ethical aspects of a clinical trial cannot beseparated from the scientific objectives.Segregation of ethical issues from the full range ofstudy design components demonstrates a flaw in understanding the fundamental nature ofresearch involving human subjects.Compartmentalization of ethical issues is inconsistentwith a well-run trial.Ethical and scientific considerations are intertwined(p.4).1.2 The Discipline of StatisticsIn the realm of clinical research the discipline of Statistics(recognized here by theuse of an upper case S to differentiate it from discussion of individual statisticssuch as the average of a group of numbers)can be thought of as an integrateddiscipline that is important in all of the following associated activities 2:Identifying a research question that needs to be answered;Deciding upon the design of the clinical trial,the methodology that will beemployed,and the numerical information(data)that will be collected;Presenting the design,methodology,and data to be collected in a Study Pro-tocol.This study protocol specifies the manner of data collection and addressesall methodological considerations necessary to ensure the collection of optimumquality data for subsequent statistical analysis.21Setting the Scene Identifying the statistical techniques that will be used to describe and analyzethe data in the study protocol(or an associated Statistical Analysis Plan that iswritten in conjunction with the study protocol);Describing and analyzing the data to evaluate whether there is compellingevidence that the drug is safe and effective.Presenting the results of a clinical study to a regulatory agency in a clinicalstudy report and presenting the results to the clinical community in conferencepresentations and journal publications.The mathematical calculations that are done while designing a clinical trial(determining the chosen sample size)and after the collection and organization ofcategorical and numerical representations of biological information(more com-monly known simply as data)are straightforward and by far the easiest aspect ofthe discipline of Statistics.The more difficult aspects are the appropriate imple-mentation of the art and science of the discipline,difficulties that argue powerfullyand persuasively for the involvement of statisticians from alpha to omega,i.e.,throughout the entire process of conceptualizing,designing,conducting,analyz-ing,interpreting,and reporting a clinical trial.It was noted in the previous section that everyone involved in conductingclinical trials can benefit from a conceptual understanding of the discipline ofStatistics.The same is true for all health professionals involved in pharmaco-therapy.Physicians who,in consultation with their patients,make decisions toimplement a therapeutic intervention(e.g.,pharmacotherapy,chemotherapy,radiotherapy,surgery)need a keen understanding of benefit-risk assessments sincethey directly influence the choice of intervention.They need to be able to readmedical journals and other legitimate sources of empirical research findings,assessthe quality of the information provided,and decide on a case-by-case basiswhether the course of action of interest is likely to have a favorable benefit-riskprofile for a given patient.As Katz 3 observed,All of the art and all of thescience of medicine depend on how artfully and scientifically we as practitionersreach our decisions.The art of clinical decision-making is judgment,an even moredifficult concept to grapple with than evidence.1.3 Generalizing Information Gained from a Clinical TrialWhile precise knowledge of the data obtained from the subjects participating in aclinical trial is vital in making go/no-go decisions within a drugs clinicaldevelopment program(i.e.,whether to progress to the next step in the program),the ultimate goal of conducting a series of related clinical trials is to form aneducated estimate of what is likely to occur in the general population with thedisease of interest should the drug be approved for marketing by one or moreregulatory agencies and then prescribed to these patients.1.2The Discipline of Statistics3This book uses a central example of developing a new drug to treat high bloodpressure,or hypertension.Such drugs are called antihypertensives.This examplehas been chosen for three reasons:Everyone has a blood pressure,which is measured in various circumstances(e.g.,visits to doctors offices and when applying for a new life insurancepolicy);The units of measurement,millimeters of mercury(mmHg),are the sameworldwide(which is not the case for other measurements such as cholesterollevels);High blood pressure is a clinical condition of great concern in many geographicregions.In the United States,for example,around 33%of adults have high bloodpressure,a major risk factor for heart disease,stroke,congestive heart failure,andkidney disease 4.It is simply not possible for all of these tens of millions ofindividuals to participate in trials conducted during a new antihypertensivesclinical development program.Therefore,on the basis of the Phase III trialsconducted at the end of the program,in which(only)several thousand subjects arelikely to have received the drug,it is incumbent upon the Sponsors statisticiansand clinicians to provide their best estimate of how safe the drug would be,andwhat degree of benefit it would bring to the general population of hypertensivesshould it be approved.Computations involving the actual blood pressure data collected in the trialsmust be used in the calculation of this estimation,which must then be presented incomplex regulatory documents that will be scrutinized by regulatory reviewers.One question therefore is:How do Sponsors collect and then analyze and interpretdata from a small subject sample in a clinical trial to enable the best possiblegeneralization to an extremely large number of individuals?And a second questionis:How do the regulatory reviewers evaluate the documents provided to them bythe Sponsor and hence decide the drugs regulatory fate?The answer to bothquestions is that they use the discipline of Statistics.1.4 Blood PressureBlood pressure measurement is a component of virtually all clinical trials.In themajority of cases it is a safety measure.Drugs for all diseases except hypertensionare not supposed to affect blood pressure,and,since blood pressure is such animportant parameter,we need to make sure that the drug is not changing(raisin